Autonomous Car Progress

[daktari]

https://money.usnews.com/investing/news/articles/2021-10-19/exclusive-biden-to-tap-no-2-official-to-head-us-auto-safety-agency-source

Take a look at Ms. Cummings twitter feed to get a sense of how she thinks of Tesla (and other AV companies). Not sure if it belongs in this thread, but an interesting choice.

I have a lot of respect for (fighter) pilots. They have some knowledge about attentiveness, machine/human interface, workloads, stress and situational awareness, risks and safety. And boredom, as a lot of flying is boring.
The arguments she tweets are valid IMO.

 

[EVNow]

Take a look at Ms. Cummings twitter feed to get a sense of how she thinks of Tesla (and other AV companies). Not sure if it belongs in this thread, but an interesting choice.

The choice depends on how she lobbied for the post. Her background or Tesla FSD comments would have no bearing.

 

[diplomat33]

Aurora has a nice little short video about how their perception and sensor fusion work:

 

[powertoold]

First Zeekr 001s are being delivered within the next 2 weeks. Now they have this Zeekr Assisted Drive (ZAD), which is supposedly the Mobileye SuperVision Full Stack of awesome. ZAD will be able to do door to door L2, just like all the Mobileye autonomous driving videos on YouTube. I can't wait to see it respond to traffic controls like AP, make left and right turns like fsd beta! (/s):

ZEEKR 001 Goes Into Production

The most anticipated electric vehicle of the year has entered into production. The first production-ready ZEEKR 001 rolled off the production line earlier today at a special ceremony in ZEEKR’s Intelligent Factory in Ningbo, China. THE INTELLIGENT FACTORY // It corrects itself. If required...

www.webwire.com

 

[diplomat33]

First Zeekr 001s are being delivered within the next 2 weeks. Now they have this Zeekr Assisted Drive (ZAD), which is supposedly the Mobileye SuperVision Full Stack of awesome. ZAD will be able to do door to door L2, just like all the Mobileye autonomous driving videos on YouTube. I can't wait to see it respond to traffic controls like AP, make left and right turns like fsd beta! (/s):

ZEEKR 001 Goes Into Production

The most anticipated electric vehicle of the year has entered into production. The first production-ready ZEEKR 001 rolled off the production line earlier today at a special ceremony in ZEEKR’s Intelligent Factory in Ningbo, China. THE INTELLIGENT FACTORY // It corrects itself. If required...

www.webwire.com

The Zeekr001 have all the SuperVision hardware but ZAD is basic AP. The full SuperVision software stack that is "L2 door to door" will come later via OTA in 2022.

 

[powertoold]

The Zeekr001 have all the SuperVision hardware but ZAD is basic AP. The full SuperVision software stack that is "L2 door to door" will come later via OTA in 2022.

Do you have a source for the 2022 part?

Do we even know what features are included in ZAD packages? What ever happened to traffic lights or stop signs?

 

[diplomat33]

Do you have a source for the 2022 part?

https://twitter.com/x/status/1450510583854403588

Do we even know what features are included in ZAD packages? What ever happened to traffic lights or stop signs?

ZAD-Standard Basic Intelligent Driving Assistance System (standard):

• AP Active Cruise Control at full speed
• Adaptive Cruise with queuing function
• AEB forward collision mitigation (with vehicle, pedestrian, bicycle, animal recognition)
• LKA lane center keeping assist
• BSD blind spot monitoring system
• DOW door open early warning system
• Collision warning after CMSR
• LCA automatic lane change warning assistance
• DPS driver status monitoring
• Speeding reminder
• Capacitive steering wheel release monitoring

ZAD-Advanced Enhanced Intelligent Driving Assistance System (optional)
No features listed yet

ZAD-Ultimate fully intelligent driving assistance system (optional)
No features listed yet

Traffic light and stop sign detection might be in ZAD-Advanced but it will definitely be in ZAD-Ultimate. ZAD-Ultimate is the full SuperVision L2 door to door package.

极氪官网

zeekrlife.com

 

[powertoold]

Cool but that twitter source isn't reliable lol

 

[Dutchy Ron]

Why this Amazon-owned company is bringing its autonomous vehicles to Seattle​

Why this Amazon-owned company is bringing its autonomous vehicles to Seattle

The Amazon-owned autonomous car company is expanding into Seattle. Here's a look under the hood at how this type of tech works.

www.popsci.com

Zoox, the self-driving car outfit, is expanding to Seattle and its rainy weather.

 

[Spuzzz]

I have a lot of respect for (fighter) pilots. They have some knowledge about attentiveness, machine/human interface, workloads, stress and situational awareness, risks and safety. And boredom, as a lot of flying is boring.
The arguments she tweets are valid IMO.

respect for fighter pilots is fine, but really….read her tweets. I couldn’t believe how many of them were overtly antiTesla. And not just about autonomous driving. She seems to be anti-lithium batteries too. It sure looks like she is obsessed with it…not a ton of other content besides Tesla and autonomous (eg Waymo) development bashing. Sprinkle in a little military software criticism and you pretty much get the picture.

 

[hgmichna]

LKA lane center keeping assist

I'd like them (and us) to avoid any reference to lane centering. Nobody wants the car to be in the center of the lane. Let me give examples for a two-lane country road.

If there is no other traffic far and wide, all risk comes from the side of the road. Safest would be to drive not in the center of the lane, but in the center of the road. But for a number of reasons we will be content with the car being in its lane, but near the centerline.

If a truck comes towards me, straddling the centerline, I want my car to be far away from it, i.e. close to the road's edge. Same while I'm being overtaken.

On a multi-lane freeway I would still like the car to move away from the center of the lane to wherever there is more free space. For example, while overtaking a truck while the other side is clear, I would like my car to move away from the truck while still staying in the lane.

So I think everybody should drop lane-centering and use "lane-keeping" or some other appropriate expression instead. Lane-centering is far from the desired autonomous driving.

 

[Dutchy Ron]

What does Saudi Arabia’s autonomous vehicle agenda mean to the rest of the world? | Computer Weekly

If all goes according to plan, within a few years, one million people will live in Saudi Arabia’s ‘city of the future’ – and every one of them will depend on a fleet of self-driving cars.

www.google.nl

What does Saudi Arabia’s autonomous vehicle agenda mean to the rest of the world?​

The team will also develop better sensor technology, which it sees as the main obstacle to achieving full automation. Lidar, the sensor technology used by Tesla and many other automobile manufacturers, has several shortcomings, including a limited range and poor performance in sandstorms and heavy rain. Because the project leaders consider these limitations even bigger showstoppers than the current state of AI, they have chosen to focus mostly on developing radar-based sensors.

 

[diplomat33]

The team will also develop better sensor technology, which it sees as the main obstacle to achieving full automation. Lidar, the sensor technology used by Tesla and many other automobile manufacturers, has several shortcomings, including a limited range and poor performance in sandstorms and heavy rain. Because the project leaders consider these limitations even bigger showstoppers than the current state of AI, they have chosen to focus mostly on developing radar-based sensors.

The article gets a few facts wrong. Tesla does not use lidar on any of their production vehicles (A few Tesla vehicles have sometimes been spotted with lidar but those are special test vehicles using lidar to help train vision, they are not production vehicles). In fact, considering what a big deal Elon has made about not using lidar, this seems like really poor fact checking to let something like that slip through.

I would also disagree that sensors are the main obstacle to achieving full automation. Companies like Waymo and Aurora have built their own in-house lidar that outperforms the off-the-shelf lidar. Also, we have HD cameras, HD lidar and HD radar now that can certainly collect all the data needed for full automation. AVs with cameras, radar and lidar collect a ton of data on their environment. The problem is not collecting data. The challenge is what you do with the perception data. The real challenge is the prediction and planning to make the right driving decisions. That is the main obstacle for achieving full automation IMO.

 

[stopcrazypp]

I'd like them (and us) to avoid any reference to lane centering. Nobody wants the car to be in the center of the lane. Let me give examples for a two-lane country road.

If there is no other traffic far and wide, all risk comes from the side of the road. Safest would be to drive not in the center of the lane, but in the center of the road. But for a number of reasons we will be content with the car being in its lane, but near the centerline.

If a truck comes towards me, straddling the centerline, I want my car to be far away from it, i.e. close to the road's edge. Same while I'm being overtaken.

On a multi-lane freeway I would still like the car to move away from the center of the lane to wherever there is more free space. For example, while overtaking a truck while the other side is clear, I would like my car to move away from the truck while still staying in the lane.

So I think everybody should drop lane-centering and use "lane-keeping" or some other appropriate expression instead. Lane-centering is far from the desired autonomous driving.

That's all well and good, but I presume the software was designed to keep the car centered by default (which is a challenge in itself already for a lot of them, many just ping-pong between lanes), and the more advanced reactions you describe may not exist at all (may lead to things like phantom braking or undesired drifts in the lane).

 

[eli_]

The article gets a few facts wrong. Tesla does not use lidar on any of their production vehicles (A few Tesla vehicles have sometimes been spotted with lidar but those are special test vehicles using lidar to help train vision, they are not production vehicles). In fact, considering what a big deal Elon has made about not using lidar, this seems like really poor fact checking to let something like that slip through.

I would also disagree that sensors are the main obstacle to achieving full automation. Companies like Waymo and Aurora have built their own in-house lidar that outperforms the off-the-shelf lidar. Also, we have HD cameras, HD lidar and HD radar now that can certainly collect all the data needed for full automation. AVs with cameras, radar and lidar collect a ton of data on their environment. The problem is not collecting data. The challenge is what you do with the perception data. The real challenge is the prediction and planning to make the right driving decisions. That is the main obstacle for achieving full automation IMO.

The hardest part is implementing a counterfactual policy. Not just predicting what other cars are going to do, but predicting the effect your behavior will have on the surrounding vehicles. I haven't heard of anyone who has a good solution to this, you need to have a model of human driving behavior. You can't just drive around like a robot.

 

[EVNow]

Why this Amazon-owned company is bringing its autonomous vehicles to Seattle​

Why this Amazon-owned company is bringing its autonomous vehicles to Seattle

The Amazon-owned autonomous car company is expanding into Seattle. Here's a look under the hood at how this type of tech works.

www.popsci.com

Zoox, the self-driving car outfit, is expanding to Seattle and its rainy weather.

I read the article twice. Still can't make out what they are doing except some mapping ... though they say ...

“We actually first sent a Highlander to Seattle in late 2019,” says Jesse Levinson, the company’s co-founder and CTO. With these SUVs, the company can use the sensors on them—devices like lidar units and cameras—to create highly-detailed maps of the urban environment, a key prerequisite for a self-driving car. “We were able to build a map of downtown Seattle very quickly, and then on our next trip we were able to drive fully autonomously in downtown Seattle,” Levinson says. “Daytime, night time, and in the rain, and it worked really, really well.”

“We did that very quietly,” he adds.

So, why exactly are they taking 2 more years to .... just do more mapping ? Why not actual offer beta test rides. Something just doesn't add up.

 

[eli_]

I read the article twice. Still can't make out what they are doing except some mapping ... though they say ...



So, why exactly are they taking 2 more years to .... just do more mapping ? Why not actual offer beta test rides. Something just doesn't add up.

They probably sent some to Seattle in 2019 to court Amazon into acquiring them, show them cool looking LIDAR maps of the Amazon HQ.

 

[EVNow]

They probably sent some to Seattle in 2019 to court Amazon into acquiring them, show them cool looking LIDAR maps of the Amazon HQ.

I can't understand why "journalists" don't ask proper followup questions and report the truth. They are like stenographers - just printing out ads for the company.

 

[diplomat33]

I read the article twice. Still can't make out what they are doing except some mapping ... though they say ...

So, why exactly are they taking 2 more years to .... just do more mapping ? Why not actual offer beta test rides. Something just doesn't add up.

They are not taking 2 more years just to do more mapping. They finished mapping. They are taking 2 years to work on the FSD and validate that it is safe enough before deployment. My guess is that it is taking them 2 years because the FSD is not good enough for actual deployment yet. For comparison, last year, Waymo reported 628,838 autonomous miles to the CA DMV with a safety disengagement rate of 1 per 29,945 miles. Zoox reported only 102,521 autonomous miles and a safety disengagement rate of only 1 per 1,627 miles. So Zoox is pretty far IMO from the level of Waymo. Yes, Zoox can drive "pretty well" right after mapping but that is a far cry from being able to actually deploy a safe and reliable ride-hailing service.

On a side note, that is why Waymo's driverless ride-hailing is impressive even though it is limited. There are plenty of AV companies that can map an area and have a few AVs that work "pretty good" but can't deploy any public ride-hailing. The fact that Waymo has been able to actually deploy driverless ride-hailing, albeit limited, is quite a feat.

 

[EVNow]

They are taking 2 years to work on the FSD and validate that

But according to them ...

So, after it "worked really, really well" they started working on the FSD ?

“We were able to build a map of downtown Seattle very quickly, and then on our next trip we were able to drive fully autonomously in downtown Seattle,” Levinson says. “Daytime, night time, and in the rain, and it worked really, really well.”

I'm sorry your skepticism doesn't extend beyond Tesla. Its very clear they are lying about something there.